Because Vedal is apparently obsessed with latency, and running the model locally.

Even if it were a human, that would require both fast typing and a fast TTS.